Summary
The objective of this research is to develop new foundations of composition in heterogeneous systems, to apply these foundations in a new generation of tools for system integration, and to validate the results in experiments using automotive and avionics System-of-Systems experimental platforms. The approach is to exploit simplification strategies: to develop theories, methods, and tools to assist in inter-layer decoupling.
Intellectual merit. The research program has three focus areas: (1) theory of compositionality in heterogeneous systems, (2) tools and tool architectures for system integration, and (3) systems/ experimental research. The project develops and deploys theories and methods for inter-layer decoupling that prevent or decrease the formation of intractable system-wide interdependences and maintain compositionality at each layer for carefully selected, essential system properties. Compositionality in tools is sought by exploring semantic foundations for model-based design. Systems/experimental research is conducted in collaboration with General Motors Global R&D (GM) and focuses on electric car platforms.
Broader impact. The project is contributing to the cost effective development and deployment of many safety and security-critical cyber-physical systems, ranging from medical devices to transportation, to defense and avionics. The participating institutions seek to complement the 30-year-old conventional curriculum in systems science with one that admits computation as a primary concept. The curriculum changes will be aggressively promoted through a process of workshops and textbook preparation.
Summary
Real-life Cyber Physical Systems (CPSs), such as autonomous vehicles and building automation systems, are monitored and controlled by networked control systems. In CPSs, the overall system dynamics emerges from the interaction among physical dynamics, computational dynamics, and communication networks. This project aims at addressing the fundamental problems in constructing CPSs caused by network uncertainties, such as time varying delay, jitter, data rate limitations, packet loss and others, by exploiting the inherent safety of passive systems.
The project will develop (1) the theoretical foundations for passivity-based design of networked control systems that provide an effective way to interconnect multiple passive systems together and preserve stability and performance in the presence of time varying delays and data dropouts and (2) model-based design processes for developing and analyzing software utilizing the compositionality and the orthogonality across design views stemming from the underlying passivity principles.
The research plan includes a software tool-chain for passivity-based design, an implementation of the passivity-based architecture on a distributed hardware platform, and experimental studies for demonstrating the approach. The project serves as an excellent example for the rich societal context and extensive interdisciplinary interactions that future computer scientists and engineers will face as CPS technology is becoming increasingly pervasive.
Summary
In the context of concurrency, programming is complicated by a number of issues that need to be addressed, such as mutual exclusion, liveness, and fairness. Petri nets (PNs) are formal models developed in Computer Science for the modeling of concurrent systems. In Control Systems, PNs have been used in the context of supervisory control (SC) of discrete event systems (DES) and powerful theoretical results have been developed. However, these results have not yet been systematically applied to Computer Science problems for which PNs were created. This research would apply SC tools to the automatic synthesis of programming code based on high-level supervisory control program specifications. SC is of interest because various desirable features of programs can be seen in terms of supervisory control specifications.
The goal of the research is to reduce programming effort by having more of the higher level requirements implemented automatically by the SC tools. While SC methods have been used to obtain control software, neither the application to the synthesis of concurrent programs nor the application of PN based SC methods have been done so far. PNs are natural models of concurrency that allow the use of a number of efficient supervisory control methods, without excluding other approaches. Another new feature of the work is the extraction of a SC specification from a specification language. This topic is of interest in order to obtain compact and easy to develop specifications, hiding the size of the underlying DES models and the technical details involved in the formulation of SC problems. Finally, the research is unique in the way it handles the SC, in an attempt to take advantage of multiple methods available in the literature.
The programming code produced by the SC tools is correct by construction and the programmer has only to manage simpler high-level specifications. The research pursued represents a fresh and novel approach to writing concurrent programs and it involves software development and research on supervisory control methods, specification languages, and code generation strategies. On one hand, the project provides a step forward towards a higher level of automation in the development of concurrent programs. On the other hand, this project will provide a platform for testing, comparing, and developing DES methods in general. Parts of the project will involve students at undergraduate level, will also benefit undergraduate students via undergraduate research and design projects. A major part of the proposed work is being carried out at a primarily undergraduate institution.